Benjamin Herrenschmidt wrote:
On Wed, 2013-12-04 at 14:45 +0100, Philippe Bergheaud wrote:
+#ifdef __LITTLE_ENDIAN__
+#define GETWORD(v) (((v)[3] << 24) + ((v)[2] << 16) + ((v)[1] << 8) +
(v)[0])
+#else
#define GETWORD(v) (((v)[0] << 24) + ((v)[1] << 16) + ((v)[2] << 8) +
(v)[3])
+#endif
#define isxdigit(c) (('0' <= (c) && (c) <= '9') \
|| ('a' <= (c) && (c) <= 'f') \
Philippe: Wouldn't it be better to just do a 32-bit load and let the
endianness be worked out
by the hardware? i.e.
#define GETWORD(v) (*(u32 *)v)
Yes, your alternative is better.
Wouldn't it narrow the scope of the macro to aligned words on POWER7?
I think that all references to GETWORD operate on aligned words anyway.
Well, xmon has to be robust ... as long as you are *certain* that even
with crap entry state it won't try to access unaligned boundaries then
go for it but we aren't looking at performance here.
Thank you Tom and Ben. We are definitely not looking at performance here.
I prefer to stay on the safe side, and leave the original patch untouched.
Philippe
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
